Categorizing continuity of care measures and why you have (probably) been doing it wrong: a methodological discussion and recommendations for researchers.

Empirical research has found associations between COC and various outcomes, including medication safety, healthcare utilization, and healthcare costs. COC thus represents an important mechanism for health policy to positively impact patient care. To craft health policy that effectively addresses COC, however, policymakers require accurate and reliable information on COC and its effects. This information is often based on routinely-collected visit data. However, researchers trying to use the right COC measures are faced with a dense jungle of (partially) incorrect categorizations, contradictory definitions and ambiguous concepts.Our paper aims to help researchers chart their path through this jungle. To do so, we first summarize and critique several approaches to the categorization of COC measures that are prominently cited in the COC literature. We find that the definition and application of COC measure categories is contradictory both within and between categorization approaches. Second, we propose a different way of categorizing COC measures, one that focuses on which patient types different measures can distinguish between. In particular, we differentiate measures by the kinds of information they encode and their mathematical properties. Finally, based on our findings, we provide some initial recommendations for researchers searching for the right COC measure for their analysis.